Weakly supervised learning for treeline ecotone classification based on aerial orthoimages and an ancillary DSM

Result code in IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216208%3A11310%2F22%3A10448917" target="_blank" >RIV/00216208:11310/22:10448917 - isvavai.cz</a>

Result on the web

<a href="https://doi.org/10.5194/isprs-annals-V-3-2022-33-2022" target="_blank" >https://doi.org/10.5194/isprs-annals-V-3-2022-33-2022</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.5194/isprs-annals-V-3-2022-33-2022" target="_blank" >10.5194/isprs-annals-V-3-2022-33-2022</a>

Result language

angličtina

Original language name

Weakly supervised learning for treeline ecotone classification based on aerial orthoimages and an ancillary DSM

Original language description

Convolutional neural networks (CNNs) effectively classify standard datasets in remote sensing (RS). Yet, real-world data are more difficult to classify using CNNs because these networks require relatively large amounts of training data. To reduce training data requirements, two approaches can be followed - either pretraining models on larger datasets or augmenting the available training data. However, these commonly used strategies do not fully resolve the lack of training data for land cover classification in RS. Our goal is to classify trees and shrubs from aerial orthoimages in the treeline ecotone of the Krkonoše Mountains, Czechia. Instead of training a model on a smaller, human-labelled dataset, we semiautomatically created training data using an ancillary normalised Digital Surface Model (nDSM) and image spectral information. This approach can complement existing techniques, trading accuracy for a larger labelled dataset while assuming that the classifier can handle the training data noise. Weakly supervised learning on a CNN led to 68.99% mean Intersection over Union ( IoU) and 81.65% mean F1-score for U-Net and 72.94% IoU and 84.35% mean F1-score for our modified U-Net on a test set comprising over 1000 manually labelled points. Notwithstanding the bias resulting from the noise in training data (especially in the least occurring tree class), our data show that standard semantic segmentation networks can be used for weakly supervised learning for local-scale land cover mapping.

Czech name

—

Czech description

—

Type

D - Article in proceedings

CEP classification

—

OECD FORD branch

10508 - Physical geography

Project

—

Continuities

S - Specificky vyzkum na vysokych skolach<br>I - Institucionalni podpora na dlouhodoby koncepcni rozvoj vyzkumne organizace

Publication year

2022

Confidentiality

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Article name in the collection

XXIV ISPRS Congress on Imaging Today, Foreseeing Tomorrow, Commission III

ISBN

—

ISSN

2194-9042

e-ISSN

2194-9050

Number of pages

6

Pages from-to

33-38

Publisher name

Copernicus Gesellschaft MBH

Place of publication

Gottingen

Event location

Nice

Event date

Jun 6, 2022

Type of event by nationality

WRD - Celosvětová akce

UT code for WoS article

000855203200006